Are you sure?

Posted Jul 31, 2009 4:34 UTC (Fri) by khim (subscriber, #9252)
In reply to: A new GCC runtime library license snag? by spitzak
Parent article: A new GCC runtime library license snag?

There are plenty of distributions that include closed-source drivers, and include GPL2/3 programs that when run will call those drivers. Nobody ever thought that was a problem.

That's because syscall interface creates demarcation line: it implements POSIX and so you can consider program which uses kernel directly. When both sides are in kernel it is a problem, when GPLed code uses non-GPLed library it's a problem too (note how QPL/GPL incompatibility was never sidestepped with "system libraries" excuse - nobody ever suggested it and RMS was even eager to point out that Misusing a GPL-covered program permanently forfeits the right to distribute the code at all.

What if the disk was partitioned, so you installed the OS first, and then installed the application second?

Courts will look at the intent. If these two discs come from separate companies then it's clearly allowed, but if they are sold as a bundle intended to be used in tandem... then the answer is probably no.

What if that was done automatically?

Probably not, but then degree of automatization matter.

How about if it was on another disk, or if the installation of the OS automatically ran a program that downloaded the app and installed it?

Again: what was the intent of all these dances, why you are doing it this way and not that way, etc.

It seems these are allowed, so disallowing the same-disk distribution is totally silly.

These are not unconditionally allowed, so distribution on the same-disk can be allowed or disallowed.

You are forgetting again that law is squishy. When you construct long series or argument you in effect construct some "chain of reasoning": if A then B, if B then C, if C then D, etc. And then you expect to use it to prove that of A then K. But lawer looks on this chain in different way: it says "if A then B - hmm, good argument, 80% probablity to withstand test in the court", then "if B then C - hmm, good argument too, 80% probability it'll be Ok in court", then you "agree" again and again and in the end you are saying "Ok, not we have the proof that if A then K" and lawer says "sorry, but we have no such proof because this chain has roughly 10% chance to withstand test in court".

Actually it's easy for geek to see how it all works if s/he ever heard about fuzzy logic. Logic related with law is fuzzy. There are no hard rules - there are probabilities. They can be bigger and smaller in different cases, but they never reach 100% and they never reach 0%. For practical purposes the fact which is supprted by ten different 99% proofs is bullet-proof (probability of failure is 1e-20 and our universe is too young to allow such a chance), but when you are using long enough chains of logic even such probabilities can be exhausted.

Also the purpose of the same-distribution clause was to prevent people from putting a big function into a closed-source "standard system component". Since the library is NOT closed source, I cannot see any plausible way that this is a problem or conflict.

Yes, that's position of FSF and it was clarified in GPLv3. But since copyright holders explicitly refused such clarification (they are not using "or later" text to show that they agree with FSF's interpretation) we can not use FSF's intent to justify such reading. The only guys who can clarify the issue are git/udev/etc copyright holders - and they are silent AFAICS...